• For computers, colors are often described in terms of the RGB channels, but that does not match up well with how humans perceive colors

Example from http://www.hclwizard.org

• Surprisingly difficult to map colors to perceptually uniform space
• Various ways to map colors; here’s a nice conceptual model of HCL space (H = hue, C = chroma, L = luminance)

There’s a nice treatment of color theory at (http://www.handprint.com/HP/WCL/wcolor.html)

HCL space is quite complicated. The colorspace package can be useful for exploring the space if you are interested. Here are slices by constant luminance, with hue mapped to angle and chrome to radius.

• Roughly 5% of people (mostly, but not exclusively, men) have color vision deficiencies
• The site http://www.color-blindness.com/coblis/coblis.html can be used to simulate color deficiencies
• The package dichromat can also be used to simulate color deficiencies (and to suggest appropriate palettes).

library(gapminder)
# from http://stat545.com/block019_enforce-color-scheme.html
jdat <- gapminder %>% 
  filter(continent != "Oceania") %>% 
  droplevels() %>% 
  mutate(country = reorder(country, desc(pop))) %>% 
  arrange(year, country)  

j_year <- 2007

jdat %>% 
  filter(year == j_year) %>% 
  ggplot(aes(x = gdpPercap, y = lifeExp, fill = country)) +
  scale_fill_manual(values = country_colors) +
  facet_wrap(~ continent) +
  geom_point(aes(size = pop), pch = 21, show.legend = FALSE) +
  scale_x_log10(limits = c(230, 63000)) +
  scale_size_continuous(range = c(1,40)) + ylim(c(39, 87))

Notce the use of scale_fill_manual(values = country_colors);

library(gapminder)
country_colors[1:20]

##          Nigeria            Egypt         Ethiopia Congo, Dem. Rep. 
##        "#7F3B08"        "#833D07"        "#873F07"        "#8B4107" 
##     South Africa            Sudan         Tanzania            Kenya 
##        "#8F4407"        "#934607"        "#974807"        "#9B4A06" 
##          Morocco          Algeria           Uganda            Ghana 
##        "#9F4D06"        "#A34F06"        "#A75106"        "#AB5406" 
##       Mozambique       Madagascar    Cote d'Ivoire         Cameroon 
##        "#AF5606"        "#B35806"        "#B75C07"        "#BA5F08" 
##     Burkina Faso           Malawi            Niger           Angola 
##        "#BE6209"        "#C2650A"        "#C5690B"        "#C96C0C"

Except in unusual circumstances, I do not recommend picking specific colors, but instead to take advantage of existing color palettes. To illustrate, we’ll fix the plot other than the colors (and color by continent, not country).

reference_plot <-
jdat %>% 
  filter(year == j_year) %>% 
  ggplot(aes(x = gdpPercap, y = lifeExp, fill = continent)) +
#  scale_fill_manual(values = continent_colors) +
  facet_wrap(~ continent) +
  geom_point(aes(size = pop), pch = 21, show.legend = FALSE) +
  scale_x_log10(limits = c(230, 63000)) +
  scale_size_continuous(range = c(1,40)) + ylim(c(39, 87))

• The default for ggplot is the function scale_color_hue(), which picks hues around the HCL wheel (with chroma and luminance fixed)

reference_plot

reference_plot + scale_fill_hue(l = 45)

reference_plot + scale_fill_hue(c = 25)

Colorbrewer, http://colorbrewer2.org, is an influential set of palettes originally selected for maps.

library(RColorBrewer)
display.brewer.all(type = "div")

display.brewer.all(type = "qual")

display.brewer.all(type = "seq")

reference_plot + scale_fill_brewer(type = "qual", palette = "Accent")

scale_color_distiller() applies the ColorBrewer color scales to continuous data

A newly popular palette comes via the viridis package, which takes an approach from MatLab.

“These color maps are designed in such a way that they will analytically be perfectly perceptually-uniform, both in regular form and also when converted to black-and-white. They are also designed to be perceived by readers with the most common form of color blindness.”"

library(viridis)

## Loading required package: viridisLite

reference_plot + scale_fill_viridis(discrete = TRUE)

If you use option = "E" in scale_fill_viridis(), you’ll get a palette suitable for those with color blindness.

library(viridis)
reference_plot + scale_fill_viridis(discrete = TRUE, option = "E")

The Wes Anderson Palettes

For all of these, remember that to use scale_color_*(), not scale_fill_*() if you want to affect the color of lines and points.

• Every plot has many non-data elements, such as background color, font, font size, and so on. You can adjust these individually with theme(), but I don’t usually bother unless I’m tweaking something for publication. (See ggThemeAssist under the RStudio Addins menu for a really useful tool, however.)
• However, there are a number of complete themes, some part of the ggplot2 package and some in the ggthemes package. The ggthemes package also include color palettes.
• You can examine the what the theme does with, for example theme_linedraw.

reference_plot

reference_plot + theme_bw()

reference_plot + theme_dark()

reference_plot + theme_classic()

library(ggthemes)
reference_plot + theme_economist() + scale_fill_economist()

reference_plot + theme_wsj()+ scale_fill_wsj()

reference_plot + theme_tufte()

reference_plot + theme_excel() + scale_fill_excel()

library(extrafont)
library(xkcd)
reference_plot + theme_xkcd() 

## Warning in theme_xkcd(): Not xkcd fonts installed! See vignette("xkcd-
## intro")

Read the xkcd vignette for font instructions

• Remember that we discussed other aspects of controlling the look of plots in Week 2.

• If you need to place several plots on the page, there are several packages you can use

  • patchwork
  • gridExtra
  • cowplots

(because this doesn’t fit anywhere else)

• If you have a factor variable involved in plotting (or a character variable that will be converted to a factor), ggplot will plot it in the order of the factor—alphabetical if nothing else
• If that’s not what you want, reordering the factor will do the trick
  • fct_reorder() to reorder by a single value
  • fct_reorder2() to reorder by two values

library(gapminder)

p <- gapminder %>% 
  filter(continent == "Asia") %>%
  mutate(country = fct_reorder2(country, .x = year, .y = lifeExp)) %>%
  ggplot(aes(x = lifeExp, y = country)) +
  geom_point()

Make a plot of the use of the name Taylor from 1986 on by sex, using points connected by lines. Experiment with different colors and themes—impress your classmates. You can find a description of the themes in the ggthemes package at https://cran.r-project.org/web/packages/ggthemes/